OpenAI & Claude Configuration

Integrate OpenAI’s GPT-4o and Anthropic’s Claude 3.5 Sonnet into your editor workflow

For developers who already have active subscriptions or credits with OpenAI or Anthropic, AI Level Scatter supports full integration with these services. Both platforms offer top-tier layout calculations and design advice.


OpenAI Configuration

1. Retrieve API Credentials

  1. Go to the OpenAI Developer Platform.
  2. Navigate to the API Keys tab and click Create new secret key.
  3. Copy the secret key (it begins with sk-).

2. Enter Credentials in Unity

  1. Select Tools → Decnet → AI Settings Manager.
  2. Change the Active Provider to OpenAI.
  3. Paste the key into the API Key input field.
  4. Select the model name:
    • gpt-4o: Best overall layout generator.
    • gpt-4o-mini: High-speed, cost-effective generation.
  5. Click Test Connection to verify connection to https://api.openai.com/v1/chat/completions.

Anthropic Claude Configuration

1. Retrieve API Credentials

  1. Go to the Anthropic Console.
  2. Navigate to API Keys and generate a new key (begins with sk-ant-).

2. Enter Credentials in Unity

  1. In the AI Settings Manager, set Active Provider to Anthropic Claude.
  2. Paste the key into the API Key field.
  3. Select your model:
    • claude-3-5-sonnet-latest: Recommended for its extreme precision in spatial layout rules and design feedback.
    • claude-3-haiku-latest: Fast response times for quick scatters.
  4. Click Test Connection to ping https://api.anthropic.com/v1/messages.

API Comparison

Feature Google Gemini OpenAI Anthropic Claude
Response Latency Low ($100-300\text{ms}$) Medium ($300-600\text{ms}$) Medium ($400-700\text{ms}$)
JSON Consistency Excellent (Native) Excellent (JSON Mode) Good (Schema via system prompt)
Setup Overhead Extremely Easy Moderate Moderate
Free Tier Available Yes No (Developer Credits) No (Developer Credits)

Custom API Endpoints (Proxies)

If you use a proxy server or custom endpoint to route your API calls:

  1. Enable the Use Custom Endpoint checkbox in the AI Settings Manager.
  2. Enter your custom base URL (e.g. https://my-custom-proxy.dev/v1/chat/completions).
  3. Ensure your proxy correctly passes the authorization headers and forwards payload structures without modifying JSON properties.

[!WARNING] Do not use untrusted proxy endpoints. Standard API requests contain your private developer keys, which could be captured by malicious intermediate servers.

[!TIP] If your company restricts external network access in the corporate firewall, look into setting up a local offline proxy or switching to the Local Offline Ollama setup to keep all data local.


This site uses Just the Docs, a documentation theme for Jekyll.